Fabrication of Silica Ultra High Quality Factor Microresonators

1. Microsphere Fabrication

1. Select a small amount (approximately 5 inches) of optical fiber, strip ~1.5″ cladding from one end and clean with either methanol or ethanol (Figure 1a, b).
2. If available, cleave the end with an optical fiber cleaver. If not available, cut with wire cutters or scissors such that ~0.5″ is left. The advantage of using an optical fiber cleaver is that it produces a very smooth, uniform cut as in Figure 1b. Excessive roughness or defects from a cut may cause uneven reflow, lowering the quality factor of the resulting spheres.
3. Expose the cleaned fiber end to 3W of CO₂ laser power focused to a ~500μm diameter spot size for ~1 second (Figure 1c, d, e). This produces spheres ~200μm in diameter; however, the size can be tuned by increasing or decreasing the diameter of the optical fiber. Slightly adjusting the laser intensity may also be necessary to reflow larger or smaller spheres.

2. Microtoroid Fabrication

1. Design and make a photomask with dark, solid circles, in the spacing and diameter of your choice. It is important to note that the toroids produced will be 25-30% smaller than the circles on the mask. For example, a solid circle with a diameter of 100 microns will produce a toroid with a diameter of approximately 75 microns. Also, it is recommended to leave at least 1-2mm of space between each circle and at least 5mm of space between arrays of circles and around the edges of the mask. Since the sample wafers must be carefully handled with tweezers, it is important to leave space for the tweezers to grip without damaging the toroids. The extra space also provides room for a tapered optical fiber to couple light into the finished devices, and allows samples to be cut into smaller arrays more easily. For this procedure, we used a mask with rows of 160 μm diameter circles ~1mm apart, with ~5mm of space between each row of circles. The finished toroids are approximately 110 μm in diameter.
2. Begin with silicon wafers with a 2 μm thick layer of thermally grown silica. Cleave the wafers to fit the desired microdisk pattern on the photolithography mask, leaving room for photoresist edge bead. Note that at the beginning of fabrication, it is usually most convenient to etch several arrays of circles on larger pieces of silicon wafers (~several cm x several cm). Larger wafers allow photolithography and BOE etching of more samples at a time, and are more easily handled with tweezers. Later, before the XeF₂ etching step, it is recommended to cleave the larger wafers into smaller arrays to allow faster, more uniform XeF₂ etching.
3. In a fumehood, thoroughly clean the wafers by rinsing with acetone, methanol, isopropanol, and deionized water. Blow the samples dry using a nitrogen or filtered compressed air gun, and place them on a hot plate set to 120°C for at least 2 minutes to dry.
4. After letting the wafers cool, place them in a flammable/solvent fumehood and expose to HMDS for 2 minutes using the vapor deposition method. A simple vapor deposition method: put a few drops of HMDS in a small 10ml beaker, and then cover the wafers and small beaker with a larger glass container to hold the vapor.
5. Place a sample on a spinner with an appropriately sized mount. Using a dropper bottle or syringe and filter, apply photoresist to the sample. Spin coat S1813 photoresist onto each sample for 5 seconds at 500rpm, followed by 45 seconds at 3000rpm. Edge bead removal is not needed if the wafer is sufficiently large so that the edge bead does not interfere with the patterning.
6. Soft bake the photoresist on a hot plate at 95°C for 2 minutes.
7. Using a UV mask aligner and the desired photomask, expose the photoresist-covered samples to a total of 80mJ/cm² of UV radiation.
8. Immerse the samples in MF-321 developer to remove the photoresist which was exposed to UV light. While developing, closely watch as the photoresist is removed from the wafer and dissolved. It is important to stir/swish the container constantly during this process to ensure the photoresist is removed uniformly. For the given parameters, the photoresist takes approximately 30 seconds to develop.
9. When most of the unwanted photoresist has dissolved in the developer, rinse the samples thoroughly under running water, gently blow dry the samples using a nitrogen or air gun, and inspect the samples with a microscope to ensure all undesired photoresist has been removed. If needed, the samples can be immersed again in developer; however, one should be careful not to overdevelop the samples as the desired photoresist patterns could also be damaged. (If the desired patterns are damaged or defective, the photoresist can be removed with acetone and steps 2.1-2.9 can be repeated again).
10. After developing, thoroughly rinse the samples in running water, gently blow dry the samples, and hard bake them on a hot plate at 110°C for 2 minutes. This step heats the photoresist above its glass transition temperature, reflowing the photoresist and partially repairing roughness which occurred during the developing process.
11. Using Teflon containers and the necessary protective equipment, immerse the samples in improved buffered oxide etchant (BOE). BOE contains HF, which etches the silica not covered by photoresist to form circular silica pads on the silicon wafer (Figure 2a-c). Improved buffered HF produces a smoother etch, minimizing roughness in the resulting silica circles. While it is possible to mix buffered HF starting with 49% HF, this can lead to highly variable results as typically only small quantities are made.
12. After approximately 15-20 minutes (depending on the patterns, sample sizes and number of samples), remove the samples from the BOE using Teflon tweezers. Carefully rinse the samples in running water. The silica has been removed when the samples become hydrophobic.
13. After etching, rinsing, and drying the samples, inspect them using an optical microscope. Check to make sure the desired patterns have been etched completely and all the unwanted silica has been removed. If needed, return the samples to the BOE for further etching. One should be careful not to overetch the samples, or the circular patterns underneath the photoresist may be damaged.
14. Once BOE etching is complete, thoroughly rinse the samples in deionized water and blow dry. If the samples are on large pieces of silicon wafer, it is also recommended to cut them (using a dicing saw or diamond scribe) into smaller pieces with individual rows of silica circles. Individual rows of circles are etched more quickly and uniformly in the XeF₂ etching step (2.16). Silicon dust produced by the cutting is removed during cleaning in the next step.
15. Remove the photoresist by rinsing with acetone, methanol, isopropanol, and deionized water, and dry the samples using a nitrogen gun and heating on a 120°C hot plate for at least 2 minutes.
16. Using a XeF₂ etcher, undercut the silicon beneath the circular silica pads to form silica microdisks (Figure 2d-f). The amount etched should be approximately 1/3 of the silica circle’s size, so that the resulting microdisk’s pillar is approximately 1/3-1/2 of the total disk diameter, as determined by inspection with an optical microscope. The number of XeF₂ pulses and the duration of each pulse depends on the amount of silicon in the chamber and the type of XeF₂ etcher used.
17. After XeF₂ etching, expose the samples to a focused CO₂ laser beam at approximately 12W intensity for ~3 seconds or until a smooth toroid is formed (Figure 2g-i). Depending on the exact size of the disk and the amount of XeF₂ undercut, a slightly higher or lower intensity and exposure time may be needed to form a microtoroid. It is important that the center of the laser beam and the center of the microdisk are aligned, so that the silica microdisk will form a smooth, circular microtoroid.

3. Representative Results

The microsphere and microtoroid devices can be imaged using both optical microscopy and scanning electron microscopy (Figure 1d, e and Figure 2h, i). In all images, the uniformity of the device surface is clearly evident.

To verify that the detailed approach creates ultra-high-Q devices, we also characterized the Q factor of several devices by performing a linewidth (Δλ) measurement and calculating the loaded Q from the simple expression: Q=λ/Δλ=ωτ, where λ=resonant wavelength, ω=frequency, and τ=photon lifetime. Representative spectra of each device fabricated using the previously detailed procedures^1,9 and a comparison graph of several devices is shown in Figure 3. The quality factors of all devices are above 10 million, with the majority being above 100 million.

The spectrum of the microsphere was a single resonance, indicating that the light coupled into either the clockwise or counter-clockwise propagating optical mode. However, the spectrum of the toroid exhibited a split resonance, indicating that light coupled into both the clockwise and counter-clockwise modes simultaneously. This phenomenon occurs when there is a slight imperfection at the coupling site. By fitting the spectrum to a dual-Lorentzian, the Q factor of both modes can be determined. The split resonance phenomena can occur in both sphere and toroid resonators, but is more frequently observed in toroids as they are more susceptible to imperfections and have fewer optical modes compared to spheres.

Figure 1. Flow chart of the microsphere cavity fabrication process. a) Rendering and b) optical micrograph of a cleaned and cleaved optical fiber. c) Rendering, d) optical micrograph and e) scanning electron micrograph of a microspere resonator.

Figure 2. Flow chart of the microtoroid cavity fabrication process. a) Rendering, b) top-view optical micrograph and c) side-view scanning electron micrograph of the circular oxide pad, as defined by photolithography and BOE etching. Note the slight wedge-shape of the oxide that is formed by the BOE. d) Rendering, e) top-view optical micrograph and f) side-view scanning electron micrograph of the oxide pad after the XeF₂ etching step. Note that the oxide disk maintains the wedge-shaped periphery. g) Rendering, h) top-view optical micrograph and i) side-view scanning electron micrograph of the microtoroid cavity.

Figure 3. Representative quality factor spectra of the a) microsphere and b) microtoroid resonant cavities as determined using the linewidth measurement method. In very high Q devices, one may observe mode-splitting or a double peak, in which light reflects off a small defect and circulates in both clockwise and counterclockwise directions. c) Comparison graph showing the Q factors of several microsphere and microtoroid resonant cavities. Click here for larger figure.

Figure 4. Schematic of the CO₂ laser reflow set-up. The CO₂ laser beam (solid blue line) is reflected and then focused on the sample. It passes through the 10.6 μm / 633 nm beam combiner, which transmits 10.6 μm and reflects 633 nm. The optical column images the reflection of the sample off of the beam combiner; therefore, the image is somewhat red. A list of the necessary parts for this setup is in Table 4.

Figure 5. Incorrectly reflowed a) microsphere and b) microtoroid resonant cavities. Due to incorrect placement within the beam, the device is mal-formed. c) As a result of a poor photomask or poor lithography, the toroid is moon-shaped.